Shift

Yet another wrapper around SHIFTX+ for MD trajectories analysis but using MPI and HDF5

Prerequisites

You need, at a minimum (requirements.txt):

• Python 2.7 (only for the moment)
• NumPy
• Pandas
• MDAnalysis
• H5py
• Matplotlib
• MPI4py

And SHIFTX2 chemical shift prediction tool (http://www.shiftx2.ca/).

Installation on UNIX

I highly recommand you to install the Anaconda distribution (https://www.continuum.io/downloads) if you want a clean python environnment with nearly all the prerequisites already installed (NumPy, Pandas, Matplotlib, MPI).

In order to use hdf5 format with MPI, you need a special version compiled in parallel mode (with mpi4py).

1 . First, you need to install mpi4py

conda install mpi4py

2 . Now, remove hdf5 and h5py from Anaconda (without removing the dependencies).

conda remove hdf5 h5py --force

3 . After, download the latest version of hdf5 and h5py

wget https://support.hdfgroup.org/ftp/HDF5/current/src/hdf5-1.10.5.tar.gz # check if it is the latest version
wget https://github.com/h5py/h5py/archive/2.9.0.tar.gz

4 . Compilation of HDF5

tar -xvf hdf5-1.10.5.tar.gz
cd hdf5-1.10.5
./configure --prefix=$CONDA_PREFIX --enable-linus-lfs --with-zlib=$CONDA_PREFIX --enable-parallel --enable-shared
make
make install

5 . Installation of h5py

cd ..
tar -xvf 2.9.0.tar.gz
cd ../h5py-2.9.0
export CC=mpicc
python setup.py configure --mpi --hdf5=$CONDA_PREFIX
python setup.py build
python setup.py install

6 . For the rest, you just have to do this

conda install MDAnalysis

How-To

1 . First you need to extract all the chemical shift from MD trajectory (or PDB files). You need at least to specify a directory with some PDB files or a topology and a dcd file (don't forget to rename all the special residues like HS[EDP] to HIS in the topology file, because SHIFTX+ won't recognize them). Prediction from PDB files will use only one processor, so you don't have to use MPI. On the contrary, to predict all the chemical shift from MD trajectories I recommand you to use MPI (knowing that SHIFTX+ is not the fastest software alive), especially if they are very long (> 100000 frames). And last advice, you will need a lot of RAM because SHIFTX+ takes at minimum 700 Mb (for what, I don't know. In comparison, SHIFTX takes nothing) multiply by the number of core you use !

python shift.py -p pdb_directory # With a single core

mpiexec -np 4 python shift.py -t topology.psf -d traj.dcd # With MPI using 4 cores

Command line options

• -p/--pdb: directory with pdb files
• -t/--top: topology file (pdb, psf)
• -d/--dcd: single trajectory of list of trajectories (dcd, xtc)
• --pH: pH (default: 7.4)
• --temperature: temperature (default: 7.4)
• -i/--interval: used frames at this interval (default: 1)
• -s/--shift: shift the resid number to match with experimental data (current resid + shift) (default: 0)

2 . And finally, compare them to experimental data. It will compute the RMSD between the prediction and the experimental data and plot the secondary chemical shift along the sequence for each element (Ca, Cb, etc ...). Finally, if you want, you can plot the chemical shift distribution for each residue (long operation)(see citation). But now, that you have the HDF5 file with the data, you can play with it and run whatever analysis you want ...

python analyze.py -c obs.bmrb -h5 shiftx.hdf5 -d dssp.file

Command line options

• -c/--obs: BMRB file with experimental chemical shift
• -h5: HDF5 file with the chemical shift
• -d/--dssp: DSSP file to plot secondary structure on the plot (Default: None)
• -p/--pdb: pdb file used to put data in bfactors column (Default: None)
• --distribution: if you want to plot the chemical shift distribution for each residue (default: False)

BMRB file format

The BMRB file that contains all the experimental chemical shift values must be like that.

        1 225 SER CA C  58.40 0.0  1
        2 225 SER C  C 174.90 0.0  1
        3 226 ALA H  H   8.31 0.0  1
        4 226 ALA N  N 126.10 0.0  1
        5 226 ALA CA C  53.50 0.0  1
        6 226 ALA CB C  18.20 0.0  1
        7 226 ALA C  C 178.40 0.0  1
        8 227 ASN H  H   8.26 0.0  1
        9 227 ASN N  N 117.00 0.0  1
       10 227 ASN CA C  54.00 0.0  1
       11 227 ASN CB C  38.30 0.0  1
       12 227 ASN C  C 175.90 0.0  1
       13 228 GLU H  H   7.95 0.0  1
       14 228 GLU N  N 119.20 0.0  1
       15 228 GLU CA C  57.10 0.0  1
       16 228 GLU C  C 177.00 0.0  1

Citation

1. Beomsoo Han, Yifeng Liu, Simon Ginzinger, and David Wishart. (2011) SHIFTX2: significantly improved protein chemical shift prediction. Journal of Biomolecular NMR, Volume 50, Number 1, 43-57. doi: 10.1007/s10858-011-9478-4.
2. Robustelli, P., Stafford, K. A., & Palmer III, A. G. (2012). Interpreting protein structural dynamics from NMR chemical shifts. Journal of the American Chemical Society, 134(14), 6365-6374.

License

MIT